JP2016517094A - Systems and methods for audience targeting - Google Patents

Abstract

A method and system for audience targeting is disclosed. According to one embodiment, a computer-implemented method includes logging content consumed by a first user group. Content is categorized into a set of topics. A set of topics is mapped to multiple users within a trade zone. Based on the profile matching and according to the topic of the advertising campaign, the second user group is identified from a plurality of users.

Description

(Field)
The present disclosure relates generally to the field of computer software and systems, and in particular to systems and methods for audience targeting.

(background)
Online advertisement placement generally refers to a slot or space on a page of a website that can be used to display an advertisement with its content. Advertisers typically bid on these ad placements made available through real-time bidding (RTB) transactions such as AdX, Admel, and Pubmatic.

  This requires the bid server to link the computer hardware to the RTB transaction. The bid server then receives the bid request via the RTB transaction. A bid request occurs when a user / Internet surfer visits a website / issuer that sells its advertising space for RTB transactions. In response to receiving the bid request, the bid server has a very short period of time (generally about 50-100 ms or less) to respond to the request. Because this bid response needs to occur in a very short period of time, it is difficult to run a large model to predict which advertisements to buy and the price to pay for them.

  Traditionally, advertisers manually create simple static rules to be enforced at the time of bidding. The advertiser observes and determines which domains are available for trading. The advertiser manually selects a domain for bidding. A few days later, the advertiser receives the report and visually reviews each domain for its click-through rate (“CTR”) to determine if the advertisement was done properly. CTR refers to the ratio of the number of times a user clicks on an advertisement, given the number of times the advertisement is displayed ("impression"). Advertisers remove domains that are not working well and add new domains. This traditional approach relies on human memory and human judgment to ensure that enough domains are selected for the campaign to meet the CTR goal and to meet the periodic impression allocation. It is a trial and error process. Therefore, this conventional approach is subject to human error. In addition, advertisers typically tend to bid on domains using a single static price, so they are often paying too much for ad placement, or more valuable bids at set prices. Has not been awarded.

  Finding, understanding and targeting the best audience in a digital advertising campaign is a difficult task. Geographic targeting uses the context information of the web page where the advertisement is placed to understand and predict user behavior within the geographic area. Demand for having a high return on investment (ROI) instead of identifying and classifying users most likely to respond to advertising campaigns and thus simply identifying websites with unpredictable and unmeasurable potential Is growing. The ability to target potential customers achieves cost effective marketing and provides a precise understanding of the market for future advertising campaigns.

  In order to achieve optimal audience targeting, issues arising from modeling and data management and reduction need to be addressed. After the traditional real-time bidding process for advertising campaigns, data representing user web consumption needs to be manageable and interpretable. Traditional context and behavioral targeting may not provide a reliable prediction that identifies potential users for ad placement. Datasets for advertisers to understand the types of users who may purchase the product, and to detect and filter intelligent bots from the resulting data sets, while efficiently predicting user behavior Need to collect.

  A method and system for audience targeting is disclosed. According to one embodiment, a computer-implemented method includes logging content consumed by a first user group. Content is categorized into a set of topics. A set of topics is mapped to multiple users within a trade zone. Based on the profile matching and according to the topic of the advertising campaign, the second user group is identified from a plurality of users.

  The foregoing and other preferred features, including various novel details of element implementations and combinations, will now be more particularly described with reference to the accompanying drawings and pointed out in the claims. It will be understood that the particular methods and circuits described herein are shown by way of illustration only and not limitation. As will be appreciated by those skilled in the art, the principles and features described herein may be employed in various and numerous embodiments without departing from the scope of the present invention.

The accompanying drawings included as part of this specification illustrate the presently preferred embodiment, together with the foregoing general description and the detailed description of the preferred embodiment provided below, and the principles described herein. Is explained and taught.
FIG. 1 illustrates an exemplary system level architecture for use with the present system, according to one embodiment. FIG. 2 illustrates an exemplary flowchart for collecting content consumed by a user, according to one embodiment. FIG. 3 illustrates an exemplary flowchart for audience targeting, according to one embodiment. FIG. 4 illustrates an exemplary computer architecture that may be used for the system, according to one embodiment.

  The figures are not necessarily drawn to scale, and elements of like structure or function are generally represented by like reference numerals throughout the figures for purposes of illustration. The figures are intended only to facilitate the description of the various embodiments described herein. The figures do not illustrate all aspects of the teachings disclosed herein and do not limit the scope of the claims.

(Detailed explanation)
A method and system for audience targeting is disclosed. According to one embodiment, a computer-implemented method includes logging content consumed by a first user group. Content is categorized into a set of topics. A set of topics is mapped to multiple users within a trade zone. The logged content topics visited by the first user group are aggregated into a user-specific interest profile so as to significantly reduce the data dimension. Based on the profile matching and according to the topic of the advertising campaign, the second user group is identified from a plurality of users.

  According to one embodiment, the digital content source includes, but is not limited to, a web page from a website providing a data feed, a web page associated with a target request, a purchased website server log, a mobile application, television programming Transcripts (and other video sources), on-demand video selection, and email.

  According to one embodiment, the present systems and methods provide learning, prediction, and targeting at an individual level. According to another embodiment, the system and method provide for learning, prediction, and targeting of individual groups or individuals within a geographic area that is smaller than a trade zone. A trade zone refers to a custom defined shape with an associated data set and exhibits various characteristics regarding the audience within those shapes. Trade zone audience targeting can produce satisfactory results when placing ads against general audiences within a geographic area, but always produces optimal results when the audience characteristics are not well known and predictable. It does not bring. Thus, targeting an audience at an individual level or an individual group that may not necessarily be in the same trade zone can yield more predictable and efficient results.

According to one embodiment, the present system and method enables learning of audience characteristics and interests by collecting consumed digital content and classifying the consumed digital content into categories of interest. This is a collection of digital content (eg, web pages, video transcripts, emails, mobile applications) linked to individuals into a series of scores by topic of interest. These interest scores are aggregated to determine a score for a user group at a trade zone level or a smaller user group, for example. The resulting tabular data structure provides a very efficient way to store and process user attributes in the targeting model.
According to another embodiment, the system and method selects a particular set of users who take active action or any other performance index, determine a common interest attribute, and form a target profile outcome. The system and method dynamically learns user-successful actions such as campaign-specific profiles and clicks. In one embodiment, the desired action is logged by the user outside the execution of the digital advertising campaign. For example, an advertiser may provide a list of users who have purchased a product or participated in a promotion. In another embodiment, individual and dynamic targeting between advertising campaigns provides insights about advertising campaign performance forecasts based on profile learning. Thus, audience targeting can be done personally and / or in a combination of targeting based on geographic areas or trade zones.

  According to one embodiment, the system and method provides an individual performance index or action rather than relying on trade zone performance predictions. Based on historical offline data and / or online data of currently running advertising campaigns, the system collects data and identifies users who respond well to a particular advertising campaign. As more advertising campaigns are run, the system builds an extensive database of users and their performance indexes. In addition, this audience-based targeting has a profile that closely matches the user whose system has not been previously serviced for a particular campaign but is known to respond well to a particular advertising campaign Enables identification of individuals. Thus, the system does not utilize a retargeting mechanism that services those users who have taken action, but rather identifies new users with similar audience profiles.

  Audience targeting can be mixed with geographic or trade zone targeting to achieve highly scalable and highly predictable targeting. Individuals within the same geographic area or trade zone may have changing behavior patterns or characteristics. However, individuals within a remote geographic area may not be identified efficiently because profile matching at an individual level can be costly, time consuming, and computationally intensive. Audience targeting may be supplemented by geographic targeting, and users with similar profiles and / or behavior patterns may be grouped into smaller custom geographic areas for group targeting.

  According to one embodiment, the system and method continuously updates the identity and classification of individuals during the course of the campaign. The system further learns about users interested in advertising campaigns through responses or proactive actions (eg, clicks, coupon downloads, calls, and application fills). The learned information is used to identify other users that have not been served before, but may respond well to advertisements by user profile matching. Exemplary profile attributes include classified web content, search queries, time patterns, history of various online actions, demographics, and offline purchases. In one embodiment, profile matching performs a supervised learning model such as a gradient boost regression tree. Learning and profile matching is unique for all advertising campaigns, but similar campaigns may be used to further identify the details of the targeting profile. As an example, a luxury car retailer may desire to increase the number of test runs. The traditional behavioral targeting approach would purchase cookies from websites where users surveyed luxury cars. The system identifies users who participated in the test drive, determines their profiles (eg, men interested in golf and gourmet cuisine, low frequency clickers), and then finds other users with similar profiles.

  The system and method provides improved learning and continually refines the prediction of user behavior at an individual level. Information used for audience targeting includes content from web pages, server logs, and transcripts.

  FIG. 1 illustrates an exemplary system level architecture for use with the present system, according to one embodiment. A user 101 having a browser browses a web page hosted by the server 108. The RTB system 106 receives bids from one or more bidders 107. Bidder 107 and / or advertising system 105 selects the appropriate advertisement and places the selected advertisement for user 101 on the web page of server 108. In one embodiment, the RTB system 106 uses a potential customer's user ID or any other user-specific information to provide a log of web pages visited by the potential customer. This information is used by other systems in the data processing system 109 and stored in the data storage system to analyze individual user behavior patterns and responsiveness to advertising campaigns.

  FIG. 2 illustrates an exemplary flowchart for collecting content consumed by a user, according to one embodiment. Content consumed by users in the trade zone over time is collected in several ways. One embodiment creates an inventory of content consumed by the user from several sources (201). Exemplary inventory sources include, but are not limited to, web pages associated with impressions visible through real-time bidding transactions, purchased server logs, mobile apps, TV programming transcripts (and other video sources), and on Demand video selection. In one embodiment, content consumed by the user is stored 202 in the data storage system, allowing further use without interfering with real-time content collection, such as a real-time bidding process. As an example, a web page associated with an impression that is visible by the real-time bidding trading system 106 associates 203 stored raw content that has been stored. This memory also contains the user ID and allows the creation of a user profile. Once the raw content has been collected, processing including sorting by topic is performed (204) while preserving the context metadata. The content and classification to be processed are associated with the original source data and the user (205) and stored in the data storage system (202).

  The data processing system 109 analyzes the consumption profile data and predicts users who are likely to take the desired action. The data processing system 109 further clusters the predicted users into meaningful groups and reports a list of users for targeting to the bidder 107 based on the granular information level.

  In one embodiment, bidder 107 does not request a cookie on user 101's web browser, and thus user privacy is maintained. A user is identified by its user ID, its time of action, or any other user-specific information without requiring a cookie.

  Each time the bidder 107 confirms the user 101 and available URLs, the data processing system 109 records the time, user ID, and URL. The topic assignment process may be performed by a topic assignment server, and the data processing system 109 communicates with the server to identify the category of the user's URL. Other useful information collected by the data processing system 109 includes the amount of traffic generated by the user, the time when the user browsed the URL, the user's search query, location, and historical ad click rate.

  Another useful source of data is events that are observed on the advertiser's confirmation page. Note that the observed event may take various forms. In one embodiment, the system and method use a cookie ID and time stamp that is sent when the user performs an action. The system and method match such events with user topics / consumption times. Examples of these observed events include user clicks, desired page visits, coupon downloads, calls, filling out application forms, watching videos, initiating or completing purchases, and / or interacting with applications. Can be mentioned. In one embodiment, these observed events are captured in real time or captured and provided offline by the advertiser. These observed events provide one way to place the user in the desired and undesired targeting segments.

  In one embodiment, using the observed event data, in addition to a random selection of other users, the classifier can use the regression tree to find new cookies (or users) that may take action in the future. Built using an ensemble. The bidder will then focus on the winning advertising space for this new audience for precise targeting. According to one embodiment, the system and method uses a new audience to improve the current set of users identified by other models and refine audience targeting.

According to one embodiment, the system and method learn globally from a user. Users with common interest are clustered into groups. The system and method provide daily advertising impressions to specific interested groups of users that match campaign goals without involving events observed from the advertiser's site. This leads to very important insights for advertisers. The system and method provide advertisers with very specific but anonymous information regarding the type of user consuming the product. The consumption information includes information regarding which user is an actual consumer and not a robot. As a result, advertisers spend less money for advertising campaigns, reducing the expense spent on impressions that are not confirmed by human users.
FIG. 3 illustrates an exemplary flowchart for audience targeting, according to one embodiment. The bidding system 107 tracks the web pages visited by the user (301). Bidding system 107 may track web pages visited by the user before, after, during, or regardless of the advertising campaign. The bidding system 107 can also receive user data from another system. Web pages and tracked user behavior may be obtained from offline historical data or real-time online data from ongoing advertising campaigns. Those web pages visited by the user are categorized into topics (302). Topics may be used to attract the interest of users who consume advertisements categorized under the topic. Once categorized, topics are mapped to users within the trade zone (or geographic area) of interest of the bidder (303). If the advertising campaign to be executed targets users nationwide, the mapping may be applied to trade zones defined as nationwide. Data processing system 109 continually learns user responses (304), and the response information is used to refine user categorization and classification. The data processing cluster 109 records the ID of the responding user, the action time, and the URL (305). Data processing system 109 identifies potential users who are likely to respond well to a particular advertising campaign based on the user's profile matching according to interest profiles learned from responding users of the advertising campaign (306). )

  FIG. 4 illustrates an exemplary computer architecture that may be used for the system, according to one embodiment. The exemplary computer architecture may be used to implement one or more components described in this disclosure, including but not limited to the present system. Architecture 400 includes a system bus 420 for communicating information, and a processor 410 coupled to bus 420 for processing information. Architecture 400 further includes a random access memory (RAM) coupled to bus 420 (referred to herein as main memory) or other for storing information and instructions to be executed by processor 410. A dynamic storage device 425 is included. Main memory 425 may also be used to store temporary variables or other intermediate information during execution of instructions by processor 410. Architecture 400 may also include a read only memory (ROM) and / or other static storage device 426 coupled to bus 420 for storing static information and instructions used by processor 410.

  A data storage device 427 such as a magnetic disk or optical disk and its corresponding drive may also be coupled to the architecture 400 for storing information and instructions. The architecture 400 can also be coupled to the second I / O bus 450 via the I / O interface 430. A plurality of I / O devices may be coupled to the I / O bus 450, including a display device 443, an input device (eg, an alphanumeric input device 442 and / or a cursor control device 441).

  Communication device 440 allows access to other computers (eg, servers or clients) over a network. Communication device 440 may include one or more modems, network interface cards, wireless network interfaces, or other interfaces such as those used to couple to Ethernet, token ring, or other types of networks. A device may be included.

Systems and methods for audience targeting have been disclosed. It should be understood that the embodiments described herein are for illustrative purposes and should not be considered as limiting the subject matter of the present disclosure. Various modifications, uses, substitutions, combinations, improvements, and production methods will be apparent to those skilled in the art without departing from the scope or spirit of the invention.

Claims (46)

A computer-implemented method,
Logging and storing digital content accessed by multiple users, wherein the digital content is provided in a network;
Providing a processing system coupled to the network and receiving the digital content for further processing;
Generating an interest profile for each of at least a subset of the plurality of users based on the digital content accessed by the user;
Identifying from the subset of users a user group that has taken at least one predetermined action when accessing the digital content;
Identifying at least one target user likely to take the at least one predetermined action from the plurality of users by comparing an interest profile of a potential target user with the interest profile of the user group And a method comprising:   The computer-implemented method of claim 1, wherein the digital content includes a website.   The computer-implemented method of claim 1, wherein the digital content is provided by a video transcript, email, or mobile application.   2. The predetermined action of claim 1, wherein the predetermined actions include clicks, visiting desired pages, downloading coupons, calling, filling out an application form, watching a video, starting or completing a purchase, and interacting with an application. Computer implementation method.   The computer-implemented method of claim 1, further comprising obtaining a predetermined action for the user group from a real-time bidding (RTB) system.   The computer-implemented method of claim 1, further comprising recording the time, user ID, and URL of the user group.   7. The computer-implemented method of claim 6, further comprising deriving a set of content topics based on the time, user ID, and URL of the user group.   Further comprising obtaining an amount of traffic generated by the user group, a time when the first user group browsed the URL, a search query for the first user group, a location, and a historical ad click rate. Item 8. The computer mounting method according to Item 7.   The computer-implemented method of claim 1, further comprising aggregating the content into user interest and behavior.   The computer of claim 1, wherein the comparing step includes generating a target interest profile based on the interest profile of the user group and comparing the potential target user interest profile with the target interest profile. Implementation method.   The computer-implemented method of claim 1, further comprising: providing a bidder with a list of identified target users through the network.   The interest profile generation step includes generating an interest profile based on the user's profile attributes, the profile attributes taking web content accessed, search queries, browse time patterns, specific online actions. The computer-implemented method of claim 1, comprising at least one of a history, demographics of the user, and offline purchases by the user. A non-transitory computer readable medium having computer readable instructions stored thereon, wherein a processor is coupled to the non-transitory computer readable medium, the processor comprising:
Logging and storing digital content accessed by multiple users, wherein the digital content is provided in a network;
Generating an interest profile for each of at least a subset of the plurality of users based on the digital content accessed by the user;
Identifying from the subset of users a user group that has taken at least one predetermined action when accessing the digital content;
Identifying at least one target user likely to take the at least one predetermined action from the plurality of users by comparing an interest profile of a potential target user with the interest profile of the user group A non-transitory computer readable medium that executes the computer readable instructions to perform:   The non-transitory computer readable medium of claim 13, wherein the digital content comprises a website.   The non-transitory computer readable medium of claim 13, wherein the digital content is provided by a video transcript, email, or mobile application.   14. The predetermined action includes: click, visit desired page, download coupon, call, fill out application form, watch video, start or complete purchase, and interact with application. Non-transitory computer readable medium.   The non-transitory computer readable medium of claim 13, wherein the predetermined action of the user group is obtained from an RTB system.   The non-transitory computer readable medium of claim 13, wherein the processor executes the computer readable instructions to record the time, user ID, and URL of the user group.   19. The non-transitory of claim 18, wherein the processor executes the computer readable instructions to derive a set of topics for the content based on the time, user ID, and URL of the user group. Computer readable medium.   The processor is readable by the computer to obtain the amount of traffic generated by the user group, the time when the user group browsed the URL, the search query for the first user group, the location, and the historical ad click rate. The non-transitory computer readable medium of claim 19, wherein the non-transitory computer readable medium executes instructions.   The non-transitory computer readable medium of claim 13, wherein the processor executes the computer readable instructions to aggregate the content into user interest and behavior.   The processor generates a target interest profile based on the interest profiles of the user group and compares the interest profile of the potential target user with the target to compare interest profiles. A non-transitory computer readable medium according to claim 1.   The non-transitory computer readable medium of claim 13, wherein the processor executes the computer readable instructions to provide a bidder with a list of identified target users over a network.   In order to generate an interest profile, the processor generates an interest profile based on the user's profile attributes, which includes accessed web content, search queries, browse time patterns, specific online actions. The non-transitory computer readable medium of claim 13, comprising at least one of a history taken, a demographic of the user, and an offline purchase by the user. A computer-implemented method,
Logging content visited by the first group of users;
Categorizing the content into a set of topics;
Mapping the set of topics to multiple users in a trade zone;
Identifying a second user group from the plurality of users according to a topic of an advertising campaign based on profile matching between the second user group and the first user group.   26. The profile matching according to claim 25, wherein the profile matching is performed by matching a classifier between observed events taken by the first user group and an interest profile of the first user group. Computer mounting method.   The computer-implemented method of claim 25, further comprising learning a response of the first user group.   The first user group response includes a click, a visit to a desired page, a coupon download, a call, filling out an application form, watching a video, starting or completing a purchase, and interacting with an application. Item 28. The computer-implemented method according to Item 27.   28. The computer-implemented method of claim 27, further comprising obtaining the first user group response from a real-time bidding (RTB) system.   26. The computer-implemented method of claim 25, further comprising recording a time, user ID, and URL for the first user group.   32. The computer-implemented method of claim 30, further comprising deriving the set of content topics based on the time, user ID, and URL of the first user group.   Obtaining an amount of traffic generated by the first user group, a time at which the first user group browsed the URL, a search query for the first user group, a location, and a historical advertisement click rate; 32. The computer-implemented method of claim 31, comprising:   26. The computer-implemented method of claim 25, further comprising aggregating the content into user interest and behavior.   26. The computer-implemented method of claim 25, further comprising clustering the second set of users into groups with common interests.   26. The computer-implemented method of claim 25, further comprising providing a bidder with a list of the second set of users. A non-transitory computer readable medium having computer readable instructions stored thereon, wherein a processor is coupled to the non-transitory computer readable medium, the processor comprising:
Logging content consumed by the first user group;
Categorizing the content into a set of topics;
Mapping the set of topics to multiple users within a trade zone;
Identifying the second user group from the plurality of users based on a profile match between the second user group and the first user group according to a topic of an advertising campaign. A non-transitory computer readable medium that executes possible instructions.   37. The profile matching of claim 36, wherein the profile matching is performed by matching a classifier between an observed event taken by the first user group and an interest profile of the first user group. A non-transitory computer readable medium.   38. The non-transitory computer readable medium of claim 36, wherein the processor executes the computer readable instructions to learn the response of the first user group.   The first user group response includes a click, a visit to a desired page, a coupon download, a call, filling out an application form, watching a video, starting or completing a purchase, and interacting with an application. 40. A non-transitory computer readable medium according to Item 38.   40. The non-transitory computer readable medium of claim 39, wherein the first user group response is obtained from an RTB system.   38. The non-transitory computer readable medium of claim 36, wherein the processor executes the computer readable instructions to record a time, user ID, and URL of the first user group.   42. The non-processor of claim 41, wherein the processor executes the computer readable instructions to derive a set of content topics based on a time, user ID, and URL of the first user group. Transient computer readable medium.   The processor includes the amount of traffic generated by the first user group, the time when the first user group browsed the URL, the search query for the first user group, the location, and the historical ad click rate. 43. The non-transitory computer readable medium of claim 42, wherein the computer readable instructions are executed to obtain.   38. The non-transitory computer readable medium of claim 36, wherein the processor executes the computer readable instructions to aggregate the content into user interest and behavior.   40. The non-transitory computer readable medium of claim 36, wherein the processor executes the computer readable instructions to cluster the second set of users into groups with common interests.   38. The non-transitory computer readable medium of claim 36, wherein the processor executes the computer readable instructions to provide a bidder with a second set of users.

Images